Example datasets

The following are example datasets that you can use with Amazon Lookout for Vision.

Image segmentation datasets

Getting started with Amazon Lookout for Vision provides a dataset of broken cookies that you can use to create an image segmentation model.

For another dataset that creates an image segmentation model, see Identify the location of anomalies using Amazon Lookout for Vision at the edge without using a GPU.

Image classification dataset

Amazon Lookout for Vision provides example images of circuit boards that you can use to create an image classification model.

You can copy the images from the https://github.com/aws-samples/amazon-lookout-for-vision GitHub repository. The images are in the circuitboard folder.

The circuitboard folder has the following folders.

• train – Images you can use in a training dataset.

• test – Images you can use in a test dataset.

• extra_images – Images you can use to run a trial detection or to try out your trained model with the DetectAnomalies operation.

The train and test folders each have a subfolder named normal (contains images that are normal) and a subfolder named anomaly (contains images with anomalies).

Note

Later, when you create a dataset with the console, Amazon Lookout for Vision can use the folder names (normal and anomaly) to label the images automatically. For more information, see Creating a dataset using images stored in an Amazon S3 bucket.

To prepare the dataset images

1. Clone the https://github.com/aws-samples/amazon-lookout-for-vision repository to your computer. For more information, see Cloning a repository.

2. Create an Amazon S3 bucket. For more information, see How do I create an S3 Bucket?.

3. At the command prompt, enter the following command to copy the dataset images from your computer to your Amazon S3 bucket.

   aws s3 cp --recursive your-repository-folder/circuitboard s3://your-bucket/circuitboard

After uploading the images, you can create a model. You can automatically classify the images by adding the images from the Amazon S3 location that you previously uploaded the circuit board images to. Remember that you are charged for each successful training of a model and for the amount of time that a model is running (hosted).

To create a classification model

1. Do Creating a project (console).

2. Do Creating a dataset using images stored in an Amazon S3 bucket.

   • For step 6, choose the Separate training and test datasets tab.

   • For step 8a, enter the S3 URI for the training images you uploaded in To prepare the dataset images. For example s3://your-bucket/circuitboard/train. For step 8b, enter the S3 URI for the test dataset. For example, s3://your-bucket/circuitboard/test.

   • Be sure to do step 9.

3. Do Training a model (console).

4. Do Starting your model (console).

5. Do Detecting anomalies in an image. You can use images from the test_images folder.

6. When you're finished with the model, do Stopping your model (console).